Self-Employment as a Career Investment
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.
Ruth: We're looking at the paper: "Self-Employment as a Career Investment".
Warren: Here's the gist of the paper. Self-Employment as a Signal: Career Concerns with Hidden Firm Performance∗ Georgy Lukyanov† Konstantin Popov‡ Shubh Lashkery§ Abstract We study a stationary labour market in which risk-averse workers privately know their permanent talent and choose, period by period, between risky self-employment, whose outcomes become part of a portable public record, and firm employment, which pays a competitive wage but keeps individual performance hidden from the outside market.
Ruth: <ref:2509.01265#pg2> Because each worker decides whether to generate another public outcome or to apply to a firm, both the population holding a given record and the pool of applicants at that record are endogenous.
Warren: <ref:2509.01265#pg4> It is shown that, when the effective continuation factor is below one half, a stationary sequential competitive equilibrium exists and occupational choice follows a talent cutoff at every record; <ref:2509.01265#pg5> the equilibrium need not be unique, and an explicit example with two distinct equilibria is provided.
Ruth: <ref:2509.01265#pg5> Firm employment persists whenever it is strictly optimal for a given type at a given record.
Paper summary: Ruth: So we're looking at this paper today called "Self-Employment as a Career Investment." It’s about how risk-averse workers decide between self-employment, where their work becomes public, and firm employment, which gives them a stable wage but keeps their performance hidden.
Warren: Exactly. The authors are studying a stationary labor market where the key is that workers privately know their permanent talent and they have to choose between these two options every period. The core question they're asking is how the market should read this public record if workers themselves are deciding whether to generate another public outcome or apply for a firm job.
Iris: It sounds like the setup is really interesting because the record itself isn't just something that happens; it’s endogenous in two ways. First, the population holding a specific record is selected by every possible occupational path that can lead to it, and second, when someone applies to a firm, they aren't just coming from everyone holding that same record.
Ruth: Right. So they construct the market beliefs in two stages. First, they figure out the stationary flow of types through all the histories leading up to and retaining each record. Only after that do they condition on what someone decides to apply for a firm job at that exact moment.
Warren: That conditioning step is crucial because it’s where the current application decision adds another layer of selection to who actually shows up at any given record. The paper claims they find a stationary sequential competitive equilibrium when the effective continuation factor is below one half.
Iris: That condition on the continuation factor, which I think relates to how much future rewards matter compared to immediate rewards, is what allows them to establish a talent cutoff at every single record. They show that this equilibrium needs not be unique either; they actually provide an explicit example with two distinct stationary equilibria.
Ruth: Two distinct ones? That’s significant because it means the structure of how people choose their careers isn't fixed, but depends on some underlying parameters in the model. They also find that firm employment persists whenever it is strictly optimal for a given type at a given record, which leads to self-confirming choices.
Warren: Self-confirming is interesting because if you are already in a job and it’s the best choice for your talent level at that record, taking the job doesn't change the public record or your wage, so you keep making that same choice until some outside event forces you out.
Iris: And when both options are chosen at a specific record, higher-talent workers choose self-employment, and lower-talent workers apply to firms. This means the competitive wage offered to applicants is below the mean talent among all holders of that record.
Paper summary: Ruth: That's where the analysis gets really concrete for us. They show that this discount—the difference between what you expect now and what you expect later—decomposes perfectly into two things: your share of workers who stay in self-employment, and the actual talent gap between those two groups.
Warren: So, when we look at the numbers, they predict that this decomposition is exactly equal to the share of people staying self-employed plus the difference in talent between public producers and applicants. It connects a decision about career choice directly to observable metrics about talent distribution.
Iris: This structure also predicts how occupational choices actually affect the flow of new public successes and failures, which tells us how much information the market actually gets from these decisions. They're not just looking at static numbers; they're looking at dynamic flows of outcomes.
Ruth: It seems like a lot of this boils down to this: the model yields four key empirical things. We’ve got positive selection into public production, a record-specific applicant wage discount, persistent choices for opaque employment spells, and these endogenous flows of public outcomes.
Warren: And what I find most interesting from an operational standpoint is that the persistence of opaque employment predicts that if you enter a firm job and things stay stable, you won't go back to public production until something outside the model forces you out with some exit probability lambda.
Iris: That exit probability lambda is tied to how long those opaque spells last, which is geometric in duration. But we have to remember the authors’ own limitation here; they are restricted by their stationary labor market assumption, and they don't explore what happens if things change more dynamically over time.
Ruth: Right, so for now, the paper gives us a very specific picture under those stationary conditions: talent cutoffs at every record and a clear way to decompose career choice into self-employment share and talent gaps. It’s about how information flows through these different career paths.
Warren: I think the title "Self-Employment as a Career Investment" fits because it frames self-employment not just as an income source, but as a signal that, at the right talent level, is better than the hidden stability of firm employment. It’s about how you invest your future performance into being public versus keeping it private for a paycheck.
Iris: So what this means for us is that if we want to understand career choices, we have to look at the record itself and how beliefs are built up over time before looking at any single decision point. That’s the identification path they use, which is pretty solid for what they claim about those cutoffs.
Ruth: Exactly. So when you think about this paper, it’s less about finding a single optimal path and more about understanding the entire structure of equilibrium that sustains itself across all possible career paths in this market setup. That's what they found with the two distinct equilibria and those cutoff behaviors at every nonterminal record.
Conclusion: Ruth: So we’re wrapping up on "Self-Employment as a Career Investment." Basically, this paper looks at how workers choose between being self-employed and working for a firm when their performance isn't immediately visible to everyone else.
Warren: It boils down to how the public record—your track record of successes and failures—shapes your career decisions over time.
Iris: The authors show that this choice isn't random; it follows a talent cutoff at every single point on that record, which means if you have a certain level of talent, you’re going one way and if you don't, another.
Ruth: And the big finding is how they break down the benefit of making those choices. The discount to taking a risk in self-employment splits perfectly between how much time you spend doing that versus the actual gap in talent between those two groups.
Warren: So, for someone just listening, it means that when you decide whether to take a risky path or stay stable with a job, your decision isn't just about the immediate money; it’s deeply tied to where you rank relative to other people holding the same record.
Iris: Exactly. And they show that even if you choose firm employment, it stays there as long as it’s still better for your specific talent level at that moment, which makes those job spells very persistent.
Ruth: It’s a lot about how information flows through these different career paths and how much the market actually learns from those choices.
Warren: If you want to see the big picture, this is about understanding why some people stick with public work while others keep their performance hidden for a paycheck.
Toulouse School of Economics · HSE University · CREST - École Polytechnique
econ.TH
Submitted: 2025-09-01
Updated: 2026-11-15
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 89/100
The gist: Self-Employment as a Signal: Career Concerns with Hidden Firm Performance∗ Georgy Lukyanov† Konstantin Popov‡ Shubh Lashkery§ Abstract We study a stationary labour market in which risk-averse…
Key concepts
- Endogenous Record Construction
- The public record is not fixed; it is built dynamically by workers' choices. Workers decide whether to generate a new public outcome through self-employment or apply to a firm, meaning who holds which record depends on these ongoing decisions.
- Talent Cutoff
- Under certain conditions, occupational choice at any given record is determined by a specific talent threshold. Workers with talent above this threshold choose self-employment, while those below it opt for firm employment. This creates a simple structure for how people navigate different career stages.
- Applicant Wage Discount Decomposition
- When workers apply to firms at a specific record, the competitive wage they receive is lower than the mean talent of all holders of that record. This discount splits into two parts: the proportion of workers who remain self-employed and the talent gap between public producers and applicants.
Terminology
Summary
Self-Employment as a Signal: Career Concerns with Hidden Firm Performance∗ Georgy Lukyanov† Konstantin Popov‡ Shubh Lashkery§ Abstract We study a stationary labour market in which risk-averse workers privately know their permanent talent and choose, period by period, between risky self-employment, whose outcomes become part of a portable public record, and firm employment, which pays a competitive wage but keeps individual performance hidden from the outside market. p. 0 Self-Employment as a Signal:
Career Concerns with Hidden Firm Performance∗
Georgy Lukyanov† Konstantin Popov‡ Shubh Lashkery§
Abstract
We study a stationary labour market in which risk-averse workers privately know
their permanent talent and choose, period by period, between risky self-employment,
whose outcomes become part of a portable public record, and firm employment, which
pays a competitive wage but keeps individual performance hidden from the outside
market. Because each worker decides whether to generate another public outcome or to
apply to a firm, both the population holding a given record and the pool of applicants at
that record are endogenous. Self-Employment as a Signal:
Career Concerns with Hidden Firm Performance∗
Georgy Lukyanov† Konstantin Popov‡ Shubh Lashkery§
Abstract
We study a stationary labour market in which risk-averse workers privately know
their permanent talent and choose, period by period, between risky self-employment,
whose outcomes become part of a portable public record, and firm employment, which
pays a competitive wage but keeps individual performance hidden from the outside
market. Because each worker decides whether to generate another public outcome or to
apply to a firm, both the population holding a given record and the pool of applicants at
that record are endogenous. Market beliefs are therefore constructed in two stages: first
from the stationary flow of types through all histories leading to—and retaining—each
record, and only then by conditioning on the current application decision. It is shown
that, when the effective continuation factor is below one half, a stationary sequential
competitive equilibrium exists and occupational choice follows a talent cutoff at every
record; the equilibrium need not be unique, and an explicit example with two distinct
equilibria is provided. Market beliefs are therefore constructed in two stages: first
from the stationary flow of types through all histories leading to—and retaining—each
record, and only then by conditioning on the current application decision. It is shown
that, when the effective continuation factor is below one half, a stationary sequential
competitive equilibrium exists and occupational choice follows a talent cutoff at every
record; the equilibrium need not be unique, and an explicit example with two distinct
equilibria is provided. It is shown
that, when the effective continuation factor is below one half, a stationary sequential
competitive equilibrium exists and occupational choice follows a talent cutoff at every
record; the equilibrium need not be unique, and an explicit example with two distinct
equilibria is provided. Firm employment persists whenever it is strictly optimal for
a given type at a given record. Firm employment persists whenever it is strictly optimal for
a given type at a given record. At any on-path record where both occupations are
chosen, higher-talent workers select into self-employment, and the applicant wage lies
below mean talent among holders of that record; this discount decomposes exactly into
the self-employed share and the talent gap between the two groups. Firm employment persists whenever it is strictly optimal for
a given type at a given record. At any on-path record where both occupations are
chosen, higher-talent workers select into self-employment, and the applicant wage lies
below mean talent among holders of that record; this discount decomposes exactly into
the self-employed share and the talent gap between the two groups. At any on-path record where both occupations are
chosen, higher-talent workers select into self-employment, and the applicant wage lies
below mean talent among holders of that record; this discount decomposes exactly into
the self-employed share and the talent gap between the two groups. The model yields
within-record predictions for occupational choice, wages, subsequent performance, and
the duration of opaque employment spells.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The core mechanism involves endogenous record construction and belief formation
The central difficulty in the model is that the record is endogenous in two ways: first, the population holding a record is selected by every occupational path that can lead to it, and second, the current application adds a further layer of selection: firms hire from workers who apply, not everyone holding the same record. <ref:2509.01265#pg4> The analysis tracks the stationary composition of workers at every record: entry, public outcomes, survival, and decisions to retain a record through firm employment jointly determine who is present there. p. 1 2
The present paper tries to fill this gap. We track the stationary composition of workers at
every record: entry, public outcomes, survival, and decisions to retain a record through firm
employment jointly determine who is present there. For a fixed strategy σ, these densities are determined
recursively. Let as(θ) denote inflow into record s from entrants or from self-employment at a predecessor,
excluding workers already at s who retained it through firm employment.Where the paper says it
Where the paper says it
The structure of occupational choice and persistence
The first result shows that occupational choice has a simple structure under a sufficient shorthorizon condition: workers below a record-specific talent threshold choose firm employment and workers above it continue in self-employment, and a stationary equilibrium exists. <ref:2509.01265#pg8> Firm employment is selfconfirming because when it is strictly preferred, the job changes neither the public record nor the stationary wage, so a surviving worker faces the same comparison again and makes the same choice—opaque employment persists until exogenous exit. p. 2 The model then yields three qualitative implications. First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups.Where the paper says it
Where the paper says it
The derivation of equilibrium existence and cutoff behavior
The model establishes that occupational choice at each record is governed by a talent threshold under Assumption 1, which requires the effective continuation factor to satisfy γ Lemma 1 provides a uniform bound showing that Us(θ) is non-decreasing in θ at every record, and for every y > x, Qs(y) − Qs(x) ≥ (1 − 2γ)(y − x) > 0. <ref:2509.01265#pg8> Proposition 2 proves that under Assumption 1, at every record s there is a cutoff cs ∈ [0, 1] such that firm employment is optimal for θ cs. p. 9 Repeating this argument proves
the result. The proposition requires neither Assumption 1 nor a cutoff strategy.Where the paper says it
The empirical implications of the equilibrium structure
The model yields three qualitative implications: first, firm employment is selfconfirming because it leaves both the record and the stationary wage unchanged, so a surviving worker faces the same comparison again and makes the same choice—opaque employment persists until exogenous exit. <ref:2509.01265#pg9> Second, at any record where both occupations are chosen, higher-talent workers select into public production and lower-talent workers apply to firms; the competitive applicant wage therefore lies below mean talent among all holders of the record, and the discount decomposes exactly into the self-employed share and the talent gap between the two groups. p. 2 First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. Third, occupational choices determine
the flow of new public successes and failures, and hence how much information the market
receives at all. Third, occupational choices determine
the flow of new public successes and failures, and hence how much information the market
receives at all. These implications are within-record comparisons: conditional on the same public record,
an application for opaque employment should predict lower performance on a comparable
later measure, applicant wages should display a record-specific selection discount, and moves
into opaque employment should persist while the worker remains in the market.Where the paper says it
Where the paper says it
Where the paper says it
Empirical content and testable predictions
The baseline yields four closely related empirical objects: positive selection into public production, a record-specific applicant-wage discount, persistent opaque-employment choices, and endogenous flows of public outcomes. <ref:2509.01265#pg8> Within-record selection into public production suggests that conditional on the same public record, workers who remain in self-employment have higher talent than workers who apply for opaque firm employment. p. 2 First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups.Where the paper says it
Where the paper says it
Where the paper says it
Scope and extensions
The baseline results are within-record restrictions; they do not order different records or imply that a success or failure raises or lowers wages. <ref:2509.01265#pg8> Extensions such as private learning about talent, observable age, richer public histories, in-firm signals, contracts, effort, and policy interventions change the state space or the game played by firms and require the belief system to be rebuilt from the induced paths. p. 2 Talent is fixed and known by the worker, individual
performance inside firms is hidden from the outside market, and wages and strategies are
stationary. Private learning, observable career age, dated histories, public in-firm signals,
richer contracts, hidden effort, and welfare analysis all change the state space or the game
played by firms, and require additional structure; Section 6 explains what would have to be
reconstructed in each case. Relative to these literatures, the contribution of the present paper is a belief system
constructed from the endogenous stationary flow of workers through records, together with
the equilibrium analysis this construction supports: existence and talent-threshold behaviour
without assuming that successes and failures order continuation values, persistence in opaque
employment, positive selection into public production, and a record-specific applicant-wage
discount with an exact decomposition. These are conditional, within-record implications;
the model does not claim a general ordering of records or a globally signed effect of greater
transparency.Where the paper says it
Where the paper says it
Conclusion
The analysis delivers four conclusions: when future career rewards are not too strong, a stationary equilibrium exists and occupational choice follows a record-specific talent threshold; the equilibrium need not be unique, and we exhibit a specification with two distinct equilibria sustained by the complementarity between cutoffs and applicant wages. <ref:2509.01265#pg8> Where both modes are used at a record, higher-talent workers continue in public production while lower-talent workers apply to firms, so the applicant wage lies below mean talent among all holders of the record, with a discount that decomposes exactly into the self-employed share and the talent gap between the two groups. p. 2 First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. Third, occupational choices determine
the flow of new public successes and failures, and hence how much information the market
receives at all.Where the paper says it
Where the paper says it
Where the paper says it
Appendix A Technical details for stationary equilibrium
The appendix supplies the population-flow, continuity, and limiting details used in the equilibrium construction. <ref:2509.01265#pg9> Lemma 3 shows that equations (2)–(4) determine a unique collection of nonnegative integrable densities (ms)s∈S. <ref:2509.01265#pg9> For every fixed record s, there is a finite constant Cs, independent of σ, such that 0 ≤ ms(θ) ≤ Csf0(θ) for almost every θ. p. 8 A stationary sequential competitive equilibrium is a tuple (U, σ, m, G, ν, w)
satisfying the following conditions. The value function U solves (13)–(14), and the action rule σ satisfies (15) at every
record and for almost every talent type. The applicant measure and its total mass at record s are
As(dθ) =
1 − σs(θ)
ms(θ) dθ, (6)
Ds =
Z 1
0
1 − σs(θ)
ms(θ) dθ. (7)
9The logic is the familiar one from survival analysis—who is present carries information beyond what
they display—with the difference that survivorship here is chosen rather than mechanical: a worker retains
the blank record because, given the wages she faces, she prefers to.Where the paper says it
Where the paper says it
Appendix B Normalised values and two elementary wage identities
For a wage vector w, let U K s(θ;w) = (1 − γ)V K s(θ;w) be normalised lifetime utility. <ref:2509.01265#pg9> At a terminal record, U K s(θ;w) = √ws. p. 0 Self-Employment as a Signal:
Career Concerns with Hidden Firm Performance∗
Georgy Lukyanov† Konstantin Popov‡ Shubh Lashkery§
Abstract
We study a stationary labour market in which risk-averse workers privately know
their permanent talent and choose, period by period, between risky self-employment,
whose outcomes become part of a portable public record, and firm employment, which
pays a competitive wage but keeps individual performance hidden from the outside
market. Because each worker decides whether to generate another public outcome or to
apply to a firm, both the population holding a given record and the pool of applicants at
that record are endogenous. (11)
For the analysis it is convenient to normalise lifetime utility by defining
Us(θ) = (1 − γ)Vs(θ). (12)
Let the normalised value of choosing self-employment at s be
Qs(θ) = (1 − γ)θ + γ
h
θUs+ (θ) + (1 − θ)Us− (θ)
i
. Let Vs(θ) denote the expected discounted utility
of a worker with talent θ at record s. The Bellman equation is
Vs(θ) = max
u(ws) + γVs(θ),
θ + γ
h
θVs+ (θ) + (1 − θ)Vs− (θ)
i
.Where the paper says it
Where the paper says it
Where the paper says it
Appendix C A fixed-point device with forced cutoff coordinates
Lemma 8 shows that a sequence of fixed points of one of the maps constructed below is chosen with η ↓ 0. Every convergent subsequence has a limit that is a stationary sequential competitive equilibrium of the capped economy. p. 0 Market beliefs are therefore constructed in two stages: first
from the stationary flow of types through all histories leading to—and retaining—each
record, and only then by conditioning on the current application decision. It is shown
that, when the effective continuation factor is below one half, a stationary sequential
competitive equilibrium exists and occupational choice follows a talent cutoff at every
record; the equilibrium need not be unique, and an explicit example with two distinct
equilibria is provided. Market beliefs are therefore constructed in two stages: first
from the stationary flow of types through all histories leading to—and retaining—each
record, and only then by conditioning on the current application decision. It is shown
that, when the effective continuation factor is below one half, a stationary sequential
competitive equilibrium exists and occupational choice follows a talent cutoff at every
record; the equilibrium need not be unique, and an explicit example with two distinct
equilibria is provided.Where the paper says it
Where the paper says it
Appendix D A high equilibrium at every finite cap
Proposition 5 shows that for every K ≥ 2 and 0
Appendix E A low equilibrium at every finite cap
Proposition 6 shows that for every K ≥ 2 and 0 < γ < 1/9, there is an equilibrium with 0
Appendix F Two distinct infinite-record limits
Theorem 2 shows that under the assumptions of Theorem 1, the model of the paper with an unbounded public record has at least two stationary sequential competitive equilibria in pure cutoff strategies at every nonterminal record. <ref:2509.01265#pg9>
References
Abou El-Komboz L Goldbeck M 2025 Career concerns as a public good: The role of signaling for open source software development Labour Economics 97 102800 doi:10.16/j.labeco.2025.102800 Altonji JG Pierret CR 2001 Employer learning and statistical discrimination Quarterly Journal of Economics 116 313–350 doi:10. p. 3 (1999a,b) emphasise the role of the information structure, and Gibbons and Murphy
(1992) study the interaction between career concerns and explicit contracts. Employer-learning
models examine how wages respond as information about productivity accumulates (Farber
and Gibbons, 1996; Altonji and Pierret, 2001), while models of internal labour markets connect
learning to wage and promotion dynamics (Gibbons and Waldman, 1999, 2006; Waldman,
2013). Open-source production offers a related setting in which current work
becomes a portable signal: Lerner and Tirole (2002) identify career incentives and labourmarket signalling as important motives for open-source participation, Abou El-Komboz and
Goldbeck (2025) study career concerns and signalling in open-source software development,
and Belenzon and Schankerman (2015) examine motivation and the sorting of human capital
into open innovation. More generally, reputation systems use public histories to guide later
transactions (Bar-Isaac and Tadelis, 2008). Terviö (2009) studies the market failure
that arises when an early employer cannot capture the full return from discovering an
inexperienced worker’s talent; our baseline characterises private occupational choice and
applicant wages, but does not determine whether the market produces too much or too
little information. Open-source production offers a related setting in which current work
becomes a portable signal: Lerner and Tirole (2002) identify career incentives and labourmarket signalling as important motives for open-source participation, Abou El-Komboz and
Goldbeck (2025) study career concerns and signalling in open-source software development,
and Belenzon and Schankerman (2015) examine motivation and the sorting of human capital
into open innovation.Where the paper says it
Where the paper says it
Where the paper says it
Dewatripont M Jewitt I Tirole J 1999a The economics of career concerns, part I: Comparing information structures Review of Economic Studies 66 183–198 doi:1. p. 3 The career-concerns literature shows how market learning from performance creates
implicit incentives: Holmström (1999) develops the central dynamic mechanism, Dewatripont
et al. (1999a,b) emphasise the role of the information structure, and Gibbons and Murphy
(1992) study the interaction between career concerns and explicit contracts. Terviö (2009) studies the market failure
that arises when an early employer cannot capture the full return from discovering an
inexperienced worker’s talent; our baseline characterises private occupational choice and
applicant wages, but does not determine whether the market produces too much or too
little information. Open-source production offers a related setting in which current work
becomes a portable signal: Lerner and Tirole (2002) identify career incentives and labourmarket signalling as important motives for open-source participation, Abou El-Komboz and
Goldbeck (2025) study career concerns and signalling in open-source software development,
and Belenzon and Schankerman (2015) examine motivation and the sorting of human capital
into open innovation. (1999a,b) emphasise the role of the information structure, and Gibbons and Murphy
(1992) study the interaction between career concerns and explicit contracts. Employer-learning
models examine how wages respond as information about productivity accumulates (Farber
and Gibbons, 1996; Altonji and Pierret, 2001), while models of internal labour markets connect
learning to wage and promotion dynamics (Gibbons and Waldman, 1999, 2006; Waldman,
2013). The career-concerns literature shows how market learning from performance creates
implicit incentives: Holmström (1999) develops the central dynamic mechanism, Dewatripont
et al. (1999a,b) emphasise the role of the information structure, and Gibbons and Murphy
(1992) study the interaction between career concerns and explicit contracts. (1999a,b) emphasise the role of the information structure, and Gibbons and Murphy
(1992) study the interaction between career concerns and explicit contracts. Employer-learning
models examine how wages respond as information about productivity accumulates (Farber
and Gibbons, 1996; Altonji and Pierret, 2001), while models of internal labour markets connect
learning to wage and promotion dynamics (Gibbons and Waldman, 1999, 2006; Waldman,
2013). (1999a,b) emphasise the role of the information structure, and Gibbons and Murphy
(1992) study the interaction between career concerns and explicit contracts. Employer-learning
models examine how wages respond as information about productivity accumulates (Farber
and Gibbons, 1996; Altonji and Pierret, 2001), while models of internal labour markets connect
learning to wage and promotion dynamics (Gibbons and Waldman, 1999, 2006; Waldman,
2013). Our focus
is on the decision that comes before the outcome—whether to generate another public signal
at all, or to enter an applicant pool instead—which makes both the population at a record
and the set of applicants endogenous objects. The career-concerns literature shows how market learning from performance creates
implicit incentives: Holmström (1999) develops the central dynamic mechanism, Dewatripont
et al. The model also belongs to the broader literature on dynamic reputation, experimentation,
and career signalling, in which current choices or evolving track records shape future beliefs
and payoffs (e. , Board and Meyer-ter-Vehn, 2013; Thomas, 2019; Kaniel and Orlov, 2025). 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. Third, occupational choices determine
the flow of new public successes and failures, and hence how much information the market
receives at all. Terviö (2009) studies the market failure
that arises when an early employer cannot capture the full return from discovering an
inexperienced worker’s talent; our baseline characterises private occupational choice and
applicant wages, but does not determine whether the market produces too much or too
little information. Open-source production offers a related setting in which current work
becomes a portable signal: Lerner and Tirole (2002) identify career incentives and labourmarket signalling as important motives for open-source participation, Abou El-Komboz and
Goldbeck (2025) study career concerns and signalling in open-source software development,
and Belenzon and Schankerman (2015) examine motivation and the sorting of human capital
into open innovation. More generally, reputation systems use public histories to guide later
transactions (Bar-Isaac and Tadelis, 2008). 5Pallais (2014) shows experimentally that detailed public evaluations of inexperienced workers improve
later employment outcomes; Stanton and Thomas (2016) study outsourcing agencies that signal worker quality;
and Leung (2018) documents how employers’ earlier experiences with workers from particular countries affect
later hiring from those countries. , Board and Meyer-ter-Vehn, 2013; Thomas, 2019; Kaniel and Orlov, 2025). Onuchic and Ray (2023) study reputational incentives in the choice between solo and
collaborative projects when joint production obscures individual credit. More generally, reputation systems use public histories to guide later
transactions (Bar-Isaac and Tadelis, 2008). 5Pallais (2014) shows experimentally that detailed public evaluations of inexperienced workers improve
later employment outcomes; Stanton and Thomas (2016) study outsourcing agencies that signal worker quality;
and Leung (2018) documents how employers’ earlier experiences with workers from particular countries affect
later hiring from those countries. More generally, reputation systems use public histories to guide later
transactions (Bar-Isaac and Tadelis, 2008). 5Pallais (2014) shows experimentally that detailed public evaluations of inexperienced workers improve
later employment outcomes; Stanton and Thomas (2016) study outsourcing agencies that signal worker quality;
and Leung (2018) documents how employers’ earlier experiences with workers from particular countries affect
later hiring from those countries. The model also connects to talent discovery. Terviö (2009) studies the market failure
that arises when an early employer cannot capture the full return from discovering an
inexperienced worker’s talent; our baseline characterises private occupational choice and
applicant wages, but does not determine whether the market produces too much or too
little information. The model also belongs to the broader literature on dynamic reputation, experimentation,
and career signalling, in which current choices or evolving track records shape future beliefs
and payoffs (e. , Board and Meyer-ter-Vehn, 2013; Thomas, 2019; Kaniel and Orlov, 2025). (1999a,b) emphasise the role of the information structure, and Gibbons and Murphy
(1992) study the interaction between career concerns and explicit contracts. Employer-learning
models examine how wages respond as information about productivity accumulates (Farber
and Gibbons, 1996; Altonji and Pierret, 2001), while models of internal labour markets connect
learning to wage and promotion dynamics (Gibbons and Waldman, 1999, 2006; Waldman,
2013).Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Pruning redundant reasoning tokens
The paper studies a stationary labour market in which risk-averse workers privately know their permanent talent and choose, period by period, between risky self-employment, whose outcomes become part of a portable public record, and firm employment, which pays a competitive wage but keeps individual performance hidden from the outside market. <ref:2509.01265#pg2> The central difficulty in the model is that the record is endogenous in two ways: first, the population holding a record is selected by every occupational path that can lead to it, and second, the current application adds a further layer of selection: firms hire from workers who apply, not everyone holding the same record. p. 1 The central difficulty is that the record is endogenous, and in two distinct ways. First,
the population holding a record is selected by every occupational path that can lead to it.Where the paper says it
How the training signal is built
The analysis constructs beliefs in two stages: first from the stationary flow of types through all histories leading to—and retaining—each record, and only then by conditioning on the current application decision. <ref:2509.01265#pg4> The object Gs is the worker’s reputation at record s, which depends on equilibrium behaviour at every predecessor record that can lead to s and on earlier retention decisions at s itself. p. 6 (5)
The object Gs is the worker’s reputation at record s. It depends on equilibrium behaviour at
every predecessor record that can lead to s and on earlier retention decisions at s itself. It depends on equilibrium behaviour at
every predecessor record that can lead to s and on earlier retention decisions at s itself. In every period, a competitive fringe of risk-neutral, short-lived firms posts a wage for
applicants at each observable record. If Ds > 0, Bayes’ rule gives the applicant belief
νs(dθ) =
1 − σs(θ)
ms(θ) dθ
Ds
. (8)
Competition and free entry then imply the zero-profit wage
ws =
Z 1
0
θ νs(dθ) =
Z 1
0
θ
1 − σs(θ)
ms(θ) dθ
Z 1
0
1 − σs(θ)
ms(θ) dθ
.Where the paper says it
Where the paper says it
Where the paper says it
Where the gains come from
The key gains come from the decomposition of the record-specific discount, ds = ¯θs − ws, which is exactly Rs/Ms (pS s − ws) > 0. <ref:2509.01265#pg8> This decomposition separates the discount into the share of workers who remain in self-employment and the talent gap between public producers and applicants. p. 2 First, firm employment is selfconfirming: when it is strictly preferred, the job changes neither the public record nor the
stationary wage, so a surviving worker faces the same comparison again and makes the same
choice—opaque employment persists until exogenous exit. 4 Second, at any record where both
occupations are chosen, higher-talent workers select into public production and lower-talent
workers apply to firms; the competitive applicant wage therefore lies below mean talent
among all holders of the record, and the discount decomposes exactly into the self-employed
share and the talent gap between the two groups. At any on-path record where both occupations are
chosen, higher-talent workers select into self-employment, and the applicant wage lies
below mean talent among holders of that record; this discount decomposes exactly into
the self-employed share and the talent gap between the two groups. The model yields
within-record predictions for occupational choice, wages, subsequent performance, and
the duration of opaque employment spells.Where the paper says it
Where the paper says it
Persistence of opaque employment
Firm employment is persistent whenever it is strictly preferred, because it leaves both the record and the stationary wage unchanged—the choice is self-confirming. <ref:2509.01265#pg9> This prediction applies to a state–type pair, not to a record at which every worker becomes absorbed. p. 2 4Persistence attaches to a state–type pair rather than to a record. At one and the same record, lower types
can be absorbed into firm employment while higher types keep producing public outcomes, so persistence
and continued public-information production coexist within a record cell. Conditional
on survival, the worker therefore faces the same record, the same stationary wage ws, and
the same continuation values in the next period. The strict inequality u(ws) > Qs(θ) is
unchanged, so firm employment is again uniquely optimal.Where the paper says it
Where the paper says it
The baseline yields four closely related empirical objects: positive selection into public production, a record-specific applicant-wage discount, persistent …
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved understanding of endogenous record formation allows AI systems to model career progression based on public outcomes rather than just internal performance metrics. This enables simulation of how workers decide
whether to generate another public outcome or to apply for a firm
period by period, which is central to the model's structure. -
Enhanced ability to predict applicant behavior at any given record by conditioning beliefs on the current application decision and the stationary flow through all histories leading to that record. This allows for more nuanced predictions of how firms
observe the record and the worker’s current application
reveals information about talent. -
Development of a mechanism to calculate an exact decomposition of applicant wage discounts into two components:
the self-employed share and the talent gap between the two groups.
This provides a precise way to attribute wage differences observed in labor market data to specific behavioral choices rather than general performance changes. -
Capacity for simulating persistence in employment spells, where
firm employment persists whenever it is strictly optimal for a given type at a given record,
even when that choice is opaque. AI can model the duration of these spells as geometrically distributed with an exit probability lambda, reflecting the economic incentive to maintain opacity. -
Ability to perform within-record comparisons: predicting that
higher-talent workers select into self-employment
andlower-talent workers apply to firms
at any on-path record where both occupations are chosen. This allows for specific predictions about observable performance measures conditional on the complete public record, rather than just general changes in wages.
Abstract
This paper studies self-employment as an investment in future employment. Workers know their ability and choose between a safe wage and a risky independent project whose outcome becomes public. Employers pay the mean productivity of their applicants. In a two-period model with uniform ability and square-root utility, we establish a range of discount factors on which greater concern for future income increases independent production in both periods and lowers all employment wages. Some workers sacrifice current expected utility to produce a record, then accept employment after either outcome. The record is most valuable to a worker of intermediate ability; the highest types remain independent. We distinguish current income insurance from the cost of giving up this record opportunity. Identifying employers' recruitment cohorts reduces the payment needed to attract some applicant pools, but leaves the equilibrium unchanged on the established range.